Skip to content
AithozPM

Public Infrastructure

Road Quality Intelligence

Using AI and computer vision to detect and quantify potholes and road conditions.

  • AithozPM product
  • pilot
  • Computer vision
  • Geospatial data
  • Asset condition modelling
  • Public sector product design
Problem
Road maintenance is funded and scheduled on incomplete information. Manual inspection is slow, expensive and covers a fraction of the network, while citizen reports are unevenly distributed and describe severity in language rather than measurement. The result is that repair budgets are allocated against the loudest complaints rather than against the worst roads.
Solution
A computer vision system that reads road surface condition from ordinary vehicle-mounted camera footage. Defects are detected, classified and measured, then tied to a precise location. Repeat passes turn single observations into a deterioration trend, so the network can be scored continuously rather than surveyed occasionally.
Business impact
Maintenance planning moves from anecdote to measurement. Authorities can see the whole network rather than the inspected part of it, direct budget at the segments that are genuinely worst, and intervene while a defect is still cheap to repair.

The problem worth solving

Anyone responsible for a road network faces the same structural problem: the network is large, inspection capacity is small, and the two numbers are not close.

Manual survey is accurate and does not scale. Citizen reporting scales and is not representative — it tells you which roads have engaged residents, not which roads are failing. Both approaches describe severity in words. “Bad”, “quite bad” and “dangerous” cannot be sorted, compared across districts or trended over time.

So budgets get allocated on a mixture of complaint volume, political attention and the memory of whoever last drove the route. Meanwhile the defects that are cheapest to fix — the ones caught early, before water gets in and a crack becomes a structural failure — are exactly the ones that never make it into a report.

How the product works

Detection from ordinary footage

The system works from cameras mounted on vehicles already travelling the network — maintenance fleets, municipal vehicles, buses. No specialist survey vehicle is required, which is what makes frequent coverage economically possible in the first place.

Classification and measurement

Detected defects are classified by type and quantified rather than described. A pothole has an extent and a severity grade; a stretch of cracking has a length and a density. These are numbers, which means they can be ranked, aggregated by segment and compared between one district and another.

Location and trend

Every observation carries a precise location, so a defect is tied to a segment of road rather than to a street name. Because the same routes are driven repeatedly, the system builds a deterioration curve per segment — the difference between knowing a road is poor and knowing it is getting worse quickly.

Where the value lands

  • Coverage. The measured proportion of the network stops being limited by inspection headcount and starts being limited by where vehicles already drive, which is most of it.
  • Prioritisation. Budget is allocated against a ranked, measured condition score rather than against complaint volume.
  • Early intervention. Catching surface defects before water ingress turns them into structural ones is the single largest cost lever in road maintenance, and it depends entirely on frequent observation.
  • Defensibility. Spending decisions become evidenced. For a public body answerable for how maintenance budget was allocated, that matters as much as the allocation itself.

Status

At pilot stage. If you are responsible for a road network or for infrastructure asset condition more broadly, we would like to talk.

More solutions

Have a business problem worth solving with AI?

Let's explore it.

Start a ConversationExplore Solutions